On Thu, Jul 04, 2002 at 04:07:42PM +0200, Bruno Marchal wrote:
> [Hal Finney wrote:]
> >In general, universes are created by more than one computer
> >program.  The measure of a universe is proportional to the number of
> >computer programs which create it.
> 
> Glad you say so. This is *not* true for Schmidhuber and those who
> keep an absolute interpretation of the SSA (cf Bostrom Self-Sampling
> Assumption). Big and slow programs are discrimate away by Juergen,
> I argued this cannot be done a priori.

What Hal said is compatible with discriminating against big (but not
against slow)  programs, if you consider infinite strings as programs,
parts of which may never be read by the computer, and interpret "the 
number of computer programs which create a universe" as the proportion of 
infinite strings which output the universe, and the size of a program as 
the number of bits of the infinite string which is read.

> I don't think we can look into an "output" in that way. For the same
> reason we cannot look at your sleeping brain for seeing if you are
> dreaming about a banana. This is again linked to the 1/3 distinction.

You're right that we can't do this very well today. We can't look at a
brain from a third person perspective and determine exactly what it's
experiencing, but we should be able to do it in the future with better
technology. Even today we can get some probabilistic information about
what the person might be experiencing. We can tell whether someone is
awake or asleep, whether he is experiencing anger or calm, etc. If we
allow looking at the environment around the brain and making probabilistic
inferences, we can do even better. For example if we see a person facing
an apple with his eyes open, then we can infer that he is probably seeing
an apple. Similarly if I see an apple, I can infer that I'm probably in a 
universe where I'm facing an apple with my eyes open.

> I would say the flaws are both in the too much static view of a computer
> output programs, where I think "universe" are first person emergent on the
> running of all computations.

What is the motivation for defining "universe" this way?

The third person concept of a universe as the running of a computer
program (I agree that the running of it is a more useful concept than the
output considered statically) is useful in simplifying decision making.  
It allows you to consider the consequences of an action in each universe
independently, knowing that they do not interact with each other.

But how do you make use of your first person concept of universe? (I'm not
sure what the concept is exactly yet, but perhaps your answer to this
question will help.)

Reply via email to